Helsinki University of Technology

About: Helsinki University of Technology is a based out in . It is known for research contribution in the topics: Artificial neural network & Finite element method. The organization has 8962 authors who have published 20136 publications receiving 723787 citations. The organization is also known as: TKK & Teknillinen korkeakoulu.

Gas sensing using air-guiding photonic bandgap fibers

Abstract: We report on experimental studies of gas sensing using air-guiding photonic bandgap fibers. The photonic bandgap fibers have at one end been spliced to standard single mode fibers for ease of use and improved stability

The complex dielectric constant of snow at microwave frequencies

Abstract: The complex dielectric constant of snow has been measured at microwave frequencies. New and old snow at different stages of metamorphosis have been studied. The results indicate that the complex dielectric constant is practically independent of the structure of snow. For dry snow, the dielectric constant is determined by the density. For wet snow, the imaginary part and the increase of the real part due to liquid water have the same volumetric wetness dependence. The frequency dependence of the complex dielectric constant of wet snow is the same as that of water. A nomograph for determining the density and wetness of wet snow from its dielectric constant is given. A snow sensor for field measurement of the dielectric constant has been developed. It can be used for determining the density and the wetness of snow by a single measurement.

Manifold Mixup: Better Representations by Interpolating Hidden States.

Abstract: Deep neural networks excel at learning the training data, but often provide incorrect and confident predictions when evaluated on slightly different test examples. This includes distribution shifts, outliers, and adversarial examples. To address these issues, we propose Manifold Mixup, a simple regularizer that encourages neural networks to predict less confidently on interpolations of hidden representations. Manifold Mixup leverages semantic interpolations as additional training signal, obtaining neural networks with smoother decision boundaries at multiple levels of representation. As a result, neural networks trained with Manifold Mixup learn class-representations with fewer directions of variance. We prove theory on why this flattening happens under ideal conditions, validate it on practical situations, and connect it to previous works on information theory and generalization. In spite of incurring no significant computation and being implemented in a few lines of code, Manifold Mixup improves strong baselines in supervised learning, robustness to single-step adversarial attacks, and test log-likelihood.

A review of cardiac image registration methods

Abstract: In this paper, the current status of cardiac image registration methods is reviewed. The combination of information from multiple cardiac image modalities, such as magnetic resonance imaging, computed tomography, positron emission tomography, single-photon emission computed tomography, and ultrasound, is of increasing interest in the medical community for physiologic understanding and diagnostic purposes. Registration of cardiac images is a more complex problem than brain image registration because the heart is a nonrigid moving organ inside a moving body. Moreover, as compared to the registration of brain images, the heart exhibits much fewer accurate anatomical landmarks. In a clinical context, physicians often mentally integrate image information from different modalities. Automatic registration, based on computer programs, might, however, offer better accuracy and repeatability and save time.

Diversity and network coherence as indicators of interdisciplinarity: Case studies in bionanoscience

Abstract: The multidimensional character and inherent conflict with categorisation of interdisciplinarity makes its mapping and evaluation a challenging task. We propose a conceptual framework that aims to capture interdisciplinarity in the wider sense of knowledge integration, by exploring the concepts of diversity and coherence. Disciplinary diversity indicators are developed to describe the heterogeneity of a bibliometric set viewed from predefined categories, i.e. using a top-down approach that locates the set on the global map of science. Network coherence indicators are constructed to measure the intensity of similarity relations within a bibliometric set, i.e. using a bottom-up approach, which reveals the structural consistency of the publications network. We carry out case studies on individual articles in bionanoscience to illustrate how these two perspectives identify different aspects of interdisciplinarity: disciplinary diversity indicates the large-scale breadth of the knowledge base of a publication; network coherence reflects the novelty of its knowledge integration. We suggest that the combination of these two approaches may be useful for comparative studies of emergent scientific and technological fields, where new and controversial categorisations are accompanied by equally contested claims of novelty and interdisciplinarity.